Learn to Code Sustainably: An Empirical Study on LLM-based Green Code Generation

• DOI: 10.48550/arxiv.2403.03344
• Publication Date: 2024-03-05
• Authors: Tina Vartziotis; Ippolyti Dellatolas; George Dasoulas; Maximilian Schmidt; Florian Schneider; Tim Hoffmann; Sotirios D. Kotso...; Michael Keckeisen
• Keywords: Code (set theory); Code of practice; Empirical Research

The increasing use of information technology has led to a significant share energy consumption and carbon emissions from data centers. These contributions are expected rise with the growing demand for big analytics, digitization, development large artificial intelligence (AI) models. need address environmental impact software increased interest in green (sustainable) coding claims that AI models can lead efficiency gains. Here, we provide an empirical study on code overview practices, as well metrics used quantify sustainability awareness In this framework, evaluate auto-generated code. auto-generate codes considered produced by generative commercial language models, GitHub Copilot, OpenAI ChatGPT-3, Amazon CodeWhisperer. Within our methodology, order these propose definition code's "green capacity", based certain metrics. We compare performance capacity human-generated generated three response easy-to-hard problem statements. Our findings shed light current contribute sustainable development.